Waypoint: Online Semi-automatic Vehicle Occupancy Data Collection System

JP Chanchico, PCM Garcia, CM Festin… - … on Information and …, 2021 - ieeexplore.ieee.org
JP Chanchico, PCM Garcia, CM Festin, WM Tan
2021 International Conference on Information and Communication …, 2021ieeexplore.ieee.org
Transportation route planning requires different kinds of data, one of them being passenger
traffic data. Collecting this data is essential in deciding how to adjust the number of public
transport vehicles in a route. Most methods used to gather this, however, involves manual
surveys or the use of automatic passenger counting (APC) systems which may not be either
readily available or fully reliable in the case of computer-vision (CV) based APCs. To save
resources, we developed a web-based semi-automatic video annotation system that utilizes …
Transportation route planning requires different kinds of data, one of them being passenger traffic data. Collecting this data is essential in deciding how to adjust the number of public transport vehicles in a route. Most methods used to gather this, however, involves manual surveys or the use of automatic passenger counting (APC) systems which may not be either readily available or fully reliable in the case of computer-vision (CV) based APCs. To save resources, we developed a web-based semi-automatic video annotation system that utilizes GPS traces to find points-of-interest in the videos to reduce the total footage time needed to be reviewed by annotators while avoiding the pitfalls of a CV-based system. Test results show that it was possible to reduce the footage time to around 14.8% of the original dataset. The annotation results show that human annotators counted around 5.5 people per splice compared to the CV algorithm's results of around 2.46 per splice. Annotators also provided data regarding boarding and alighting with averages of 0.34 and 0.13 per splice. The reliability of the annotators was also seen by checking the standard deviation of the annotations with the 3 parameters having small standard deviations of around 1.31 (people count), 0.13 (boarding), and 0.06 (alighting), proving the reliability of the provided data. The system performance mostly shows that it can reliably reduce footage time that needs to be reviewed while providing reliable annotations of passenger traffic data.
ieeexplore.ieee.org
Showing the best result for this search. See all results